Soil pollution is a serious threat to human life and development. Different remedial measures are applied to soils with different levels of contamination. The degree of soil contamination in different areas is generally evaluated and categorised based on the analysis of samples. Regional soil sampling sites are generally sparse because of the cost of sampling and other factors, which makes it difficult to accurately assess the extent of regional soil contamination. In this study, a spatial classification model was established for the Di(2-ethylhexyl) phthalate (DEHP) pollution level using a Bi-directional Long Short-Term Memory (Bi-LSTM) neural network considering that the sampling information gradually diminishes with increasing distance between the sampling and prediction points. In this study, a method is proposed for the prediction of the spatial distribution of soil pollution categories based on sparse samples. We also established a model for the spatial distribution of organic pollution categories. The analysis of an actual contaminated area shows that the DEHP concentrations at different locations can be effectively predicted with the proposed method by categorising the contamination levels of specific DEHP samples. The results show that the method can be used to classify the degree of light/severe DEHP contamination. The results are in good agreement with the actual situation, verifying the validity of the method. This method is important for the rapid assessment of the spatial distribution of soil contamination levels based on sparse sampling.
Keywords: Deep learning; Phthalic acid esters; Soil contamination.
Copyright © 2021 The Authors. Published by Elsevier Inc. All rights reserved.